arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1908.06353cs.LGstat.ML
收藏

持续对抗扰动下神经网络控制策略的验证

Verification of Neural Network Control Policy Under Persistent Adversarial Perturbation

GE Research(通用电气研究院);MIT(麻省理工学院)

Yuh-Shyang Wang, Tsui-Wei Weng, Luca Daniel

摘要:

众所周知,深度神经网络对于小的对抗性扰动是脆弱的。当神经网络与物理系统在闭环中互连时,这个问题变得更加关键。在本文中,我们将展示如何结合最近的工作神经网络认证工具(主要用于静态设置,如图像分类)与鲁棒控制理论,以证明一个神经网络政策的控制回路。具体地说,我们给出了一个充分条件和算法,以确保闭环状态和控制约束满足时,持续对抗扰动是l-无穷范数有界的。我们的方法是基于找到一个正不变集的闭环动力系统,因此,我们不需要可微性或连续性的神经网络政策。随着验证结果,我们还开发了一个有效的攻击策略,神经网络控制系统,优于穷举蒙特卡洛搜索显着。我们表明,我们的认证算法工作良好的学习模型,并取得了5倍以上的结果比传统的Lipschitz为基础的方法来证明的神经网络政策的鲁棒性的推车杆控制问题。

英文摘要:

Deep neural networks are known to be fragile to small adversarial perturbations. This issue becomes more critical when a neural network is interconnected with a physical system in a closed loop. In this paper, we show how to combine recent works on neural network certification tools (which are mainly used in static settings such as image classification) with robust control theory to certify a neural network policy in a control loop. Specifically, we give a sufficient condition and an algorithm to ensure that the closed loop state and control constraints are satisfied when the persistent adversarial perturbation is l-infinity norm bounded. Our method is based on finding a positively invariant set of the closed loop dynamical system, and thus we do not require the differentiability or the continuity of the neural network policy. Along with the verification result, we also develop an effective attack strategy for neural network control systems that outperforms exhaustive Monte-Carlo search significantly. We show that our certification algorithm works well on learned models and achieves 5 times better result than the traditional Lipschitz-based method to certify the robustness of a neural network policy on a cart pole control problem.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑